Data mining techniques to extract academic behavior profiles and predict university desertion

Bedregal-Alpaca N.; Aruquipa-Velazco D.; Cornejo-Aparicio V. · RISTI - Revista Iberica de Sistemas e Tecnologias de Informacao (2020)

Resumen

The desertion of university students is a problem to which universities dedicate their efforts; a situation that requires more attention due to the demands of the accreditation processes. This research uses classification techniques, implemented with IBM SPSS Modeler, to predict possible student desertion. The differentiating factor of the proposal is to use indices, which in addition to considering a student’s academic performance, also place it within their cohort. To compare and evaluate the accuracy of the models the confusion matrix is used, the results indicate that the CHAID 1 tree model reaches an accuracy of 90.24%. It concludes that the total performance index is the most influential variable in desertion and that Data Mining Techniques are useful and effective in detecting patterns and predicting students’ academic behavior.

Instituciones: Universidad Nacional de San Agustín de Arequipa